Washington University School of MedicineDigital Commons@Becker

Open Access Publications

2015

Mental illness, poverty and stigma in India: Acase–control studyJean-Francois TraniWashington University in St Louis

Parul BakhshiWashington University School of Medicine in St. Louis

Jill KuhlbergWashington University in St Louis

Sreelatha S. NarayananPGIMER- Dr. Ram Manohar Lohia Hospital

Hemalatha VenkataramanRadboud University

See next page for additional authors

Follow this and additional works at: http://digitalcommons.wustl.edu/open_access_pubs

This Open Access Publication is brought to you for free and open access by Digital Commons@Becker. It has been accepted for inclusion in OpenAccess Publications by an authorized administrator of Digital Commons@Becker. For more information, please contact engeszer@wustl.edu.

Recommended CitationTrani, Jean-Francois; Bakhshi, Parul; Kuhlberg, Jill; Narayanan, Sreelatha S.; Venkataraman, Hemalatha; Mishra, Nagendra N.; Groce,Nora E.; Jadhav, Sushrut; and Deshpande, Smita, ,"Mental illness, poverty and stigma in India: A case–control study." BMJ Open.5,2.e006355. (2015).http://digitalcommons.wustl.edu/open_access_pubs/3763

AuthorsJean-Francois Trani, Parul Bakhshi, Jill Kuhlberg, Sreelatha S. Narayanan, Hemalatha Venkataraman,Nagendra N. Mishra, Nora E. Groce, Sushrut Jadhav, and Smita Deshpande

This open access publication is available at Digital Commons@Becker: http://digitalcommons.wustl.edu/open_access_pubs/3763

Mental illness, poverty and stigmain India: a case–control study

Jean-Francois Trani,1 Parul Bakhshi,2 Jill Kuhlberg,1 Sreelatha S Narayanan,3

Hemalatha Venkataraman,4 Nagendra N Mishra,3 Nora E Groce,5 Sushrut Jadhav,6

Smita Deshpande3

To cite: Trani J-F, Bakhshi P,Kuhlberg J, et al. Mentalillness, poverty and stigmain India: a case–controlstudy. BMJ Open 2015;5:e006355. doi:10.1136/bmjopen-2014-006355

▸ Prepublication history forthis paper is available online.To view these files pleasevisit the journal online(http://dx.doi.org/10.1136/bmjopen-2014-006355).

Received 12 August 2014Revised 13 January 2015Accepted 15 January 2015

For numbered affiliations seeend of article.

Correspondence toDr Jean-Francois Trani;jtrani@wustl.edu

ABSTRACTObjective: To assess the effect of experienced stigmaon depth of multidimensional poverty of persons withsevere mental illness (PSMI) in Delhi, India, controllingfor gender, age and caste.Design: Matching case (hospital)–control (population)study.Setting: University Hospital (cases) and NationalCapital Region (controls), India.Participants: A case–control study was conductedfrom November 2011 to June 2012. 647 casesdiagnosed with schizophrenia or affective disorderswere recruited and 647 individuals of same age, sex andlocation of residence were matched as controls at a ratioof 1:2:1. Individuals who refused consent or providedincomplete interview were excluded.Main outcome measures: Higher risk of poverty dueto stigma among PSMI.Results: 38.5% of PSMI compared with 22.2% ofcontrols were found poor on six dimensions or more.The difference in multidimensional poverty index was69% between groups with employment and income ofthe main contributors. Multidimensional poverty wasstrongly associated with stigma (OR 2.60, 95% CI 1.27to 5.31), scheduled castes/scheduled tribes/otherbackward castes (2.39, 1.39 to 4.08), mental illness(2.07, 1.25 to 3.41) and female gender (1.87, 1.36 to2.58). A significant interaction between stigma, mentalillness and gender or caste indicates female PSMI orPSMI from ‘lower castes’ were more likely to be poordue to stigma than male controls (p<0.001) or controlsfrom other castes (p<0.001).Conclusions: Public stigma and multidimensionalpoverty linked to SMI are pervasive and intertwined. Inparticular for low caste and women, it is a strongpredictor of poverty. Exclusion from employment linkedto negative attitudes and lack of income are the highestcontributors to multidimensional poverty, increasing theburden for the family. Mental health professionals needto be aware of and address these issues.

INTRODUCTIONMental health problems affect 450 millionpeople worldwide, 80% in middle-income andlow-income countries. In 2010, 2 319 000persons died of mental and behavioural

disorders.1 Mental health conditions accountfor 13% of the total burden of disease, 31% ofall years lived with disability and are one of thefour main contributors to years lived with dis-ability.2 3 Schizophrenia and bipolar disorderrepresent 7.4% and 7.0% of disability-adjustedlife years caused by mental and substance usedisorders, respectively.4 Severe mental illness(SMI) is a leading cause of disability, and thestandard prevalent biomedical care model isneither an exclusive nor a comprehensive solu-tion as it does not address the link betweenmental illness, stigma and poverty.5

While the literature on poverty, poor mentalhealth6 and disability7–9 is emerging, little hasbeen done to examine the compounding asso-ciations between experienced stigma (unfairtreatment or discrimination due to having amental health issue),10 mental illness andpoverty, especially in low-income countries. Inhigh-income countries,11 income deprivationis identified as a major risk factor for personswith mental health issues, even for commonmental disorders.12 Poor mental healthlinked to SMI has been associated withpoverty during the recent economic crisis inmiddle-income and low-income countries,

Strengths and limitations of this study

▪ There is little research on the effects of stigmaand poverty in developing settings.

▪ Lack of employment and income are major con-tributors to multidimensional poverty for personswith severe mental illness (PSMI).

▪ Intensity of multidimensional poverty is higherfor PSMI, particularly women with SMI andthose from scheduled castes/scheduled tribes/other backward castes.

▪ Stigma was operationalised through a single-itemquestion rather than a multiple-item scale, and wecould not assess reliability of this item. SMI wasdiagnosed for persons attending a public psychi-atric department; PSMI not receiving medical treat-ment might be more marginalised and at greaterrisk of poverty than those receiving healthcare.

Trani J-F, et al. BMJ Open 2015;5:e006355. doi:10.1136/bmjopen-2014-006355 1

Open Access Research

particularly India and China.13–15 People with mentaldisorders living in these countries are not only morelikely to be poorer, but also unemployed and less edu-cated.16 17 Indisputably, a better understanding of therelationship between mental illness and poverty mayyield useful knowledge to tailor public health interven-tions to complement biomedical treatment to improveoutcomes.Link and Phelan18 defined stigma as a process with five

interrelated components: discrimination through a processof separation based on negative attitudes and prejudiceresulting from labelling and cultural stereotypes of societytowards the stigmatised group leading to social, economicand political power differences. Thornicroft et al19 identifythree elements of stigma: ignorance or misinformation,prejudice and discrimination.Our paper focuses on the process of experienced dis-

crimination as the manifestation of public stigma.20 Thecongruence of self-stigma and social exclusion may leadpersons with SMIs (PSMIs) to face unfair treatment ordiscrimination and develop low self-esteem.21–24 Suchstigma may prevent mentally ill persons from improvingtheir conditions25 by creating a ‘barrier to recovery’26

and worsen their situation by pushing them into povertythrough discriminatory practices.27–29

Stigma towards PSMI resulting in discrimination30 31 ispersistent in India.32 Although the factors constitutingpoverty and discrimination linked to mental illness poten-tially can deprive persons of many resources,33 34 thedynamics of poverty, discrimination and mental healthhave not been fully addressed. The clinical literatureargues that stigma is caused by mental illness and treatingthe latter biomedically will weaken the associatedstigma.35 36 We argue instead that even treated PSMI aremore likely to be multidimensionally poor due to discrim-ination resulting from stigma.Many studies have focused on unidimensional effect of

poverty on mental health, but have not explained howstigma towards mental illness can be an aggravating con-tributor to the intensity of poverty. We aimed to estimatethe difference in incidence and intensity of povertybetween PSMI and a comparable control group using amultidimensional poverty index (MPI) to explore depriv-ation in various dimensions of life.37 Going beyondtraditional welfare economics approaches to poverty(ie, income or per capita expenditure), we explored non-monetary dimensions of poverty such as education, health,quality of shelter, food intake and political participation.We assessed differences in intensity of poverty betweenPSMI and controls and explored how these differencesvary as a function of discrimination resulting from stigma.

METHODSStudy design and settingThe primary objective was to assess differences in expos-ure to discrimination resulting from stigma and multidi-mensional poverty among cases compared with

non-psychiatrically ill controls. Between November 2011and June 2012, we carried out a case–control studybased at the Department of Psychiatry of the Dr RamManohar Lohia (RML) Hospital in New Delhi (cases)and in the neighbourhood of residence of the cases(controls) to assess the impact of stigma associated tomental illness on poverty. The Department of Psychiatryat Dr RML hospital received respectively 10 881 and19 528 new outpatients and 52 389 and 45 319 follow-upsof existing patients in 2012 and 2013. The departmentalso has a 42-bed general psychiatry and de-addictioninpatient facility for men and women. It serves patientsfrom the national Capital Region of Delhi.

ParticipantsWe defined cases as outpatients diagnosed with schizophre-nia or affective disorders by one of the 10 board-certifiedtreating psychiatrists following International Classificationof Diseases, 10th revision (ICD-10) criteria.38 Outpatientswere informed about the study, and if they consented, werereferred to researchers for written informed consent andevaluation with no further contact with those who refused.Transportation costs and a meal were provided to maximiserecruitment and reduce selection bias.We used a non-psychiatrically ill control group composed

of randomly selected individuals matching the patientsaccording to gender, age (±5 years) and neighbourhood ofresidence. Matched controls were selected by spinning apointer at the door of the case’s home and randomly select-ing one household among 30 in the pointed direction. Inthis household, a person of same age and gender with noreported history of a mental health disorder was inter-viewed. It was not possible to conduct detailed interviewsfor diagnosis of all controls due to logistics as well as stigmaof revealing mental illness. We excluded controls whenunable to obtain consent. Only two case patients were notmatched. Investigators together with the team managercontributed to sensitisation and awareness raising in theneighbourhoods of interest to maximise controls’ participa-tion rates. Consent for patients and controls adolescentbetween 13 and 18 was obtained by asking the parent orthe legal guardian of the study subjects.We conducted face-to-face interviews with all PSMI or a

caregiver during hospital visits, and controls at home. Weobtained information on demographics, socioeconomicfactors, health conditions and accessibility to services, edu-cation, employment, income, livelihoods and social partici-pation. The instrument was translated by experts intoHindi with iterative back-translation and tested in a pilotsurvey in October 2011. Investigators trained 2 experi-enced supervisors and 10 masters-level students over2 weeks on survey concepts and goals, mental illness aware-ness, interview techniques followed by review, test anddebriefing.

Sample sizeTo determine sample size, we used a matched designwith a control to case ratio of 1, the probability of

2 Trani J-F, et al. BMJ Open 2015;5:e006355. doi:10.1136/bmjopen-2014-006355

Open Access

exposure to poverty among controls of 0.22 and the cor-relation coefficient for exposure between matched casesand controls of 0.18.39 Considering the true OR for onedimension of poverty in exposed subjects relative tounexposed subjects as 2.2, we needed to enrol 205 casepatients to be able to reject the null hypothesis that thisOR equals 1 with probability of 0.9. The type 1 errorprobability associated with this test of this null hypoth-esis is 0.05. We enrolled 649 case patients to allow forsubgroup analyses including impact on poverty of dis-crimination stratified by gender, age and caste.

Efforts to minimise biasNew patients were first interviewed by a junior psychiatristwho made a provisional diagnosis and discussed detailswith a board-certified psychiatrist who then diagnosed andmanaged the case. To minimise diagnosis bias, we trainedall psychiatrists on the ICD-10 criteria. Information biaswas minimised by reviewing the questionnaire aboutexposure to poverty to ensure accuracy, completeness andcontent validity with experts and by testing it with a samplegroup of patients and families. Investigators revised thecontent for relevance to poverty in order to maximise itemappropriateness. They first defined the concept of multidi-mensional poverty and reviewed the empirical and theor-etical literature to identify the right deprivation items toinclude in the instrument they were developing. Theythen checked whether the questions covered all dimen-sions of the concept of multidimensional poverty andwhether the phrasing respectively in English and Hindiwas accurately reflecting the underlying concept of depriv-ation we were looking for in each dimension. Two expertsfamiliar with multidimensional poverty reviewed the initiallist of items and made suggestions about adding items thatwere omitted. We then organised a focus group discussion(FGD) with seven experts, psychiatrists, psychologists andsocial workers from Dr RML hospital to establish whetherthe 17 domains of poverty selected were adapted and rele-vant for the context of New Delhi and were providing acomprehensive overview of the concept. They also rankedthese domains by order of importance of deprivation.A similar focus group was organised with eight hospitaloutpatients with SMI. We finally tested the poverty ques-tionnaire with a group of 20 outpatients at theDepartment of Psychiatry at Dr RML hospital. Weprompted them with questions to check for their under-standing of poverty to identify the language they used toexplain the notion of poverty, as well as to ascertain theirunderstanding of the questions in order to make sure theinstrument’s purpose made sense to them. Finally, twoother experts revised the final version to make sure itemsillustrate the content of multidimensional poverty.40

We also carried out test–retest to test for recall biasand social desirability bias. Interviews with 71 respon-dents (both cases and controls) for test–retest reliabilitywere carried out on two occasions with a gap of10–15 days by the same enumerator to check to whatdegree a given respondent provided same responses for

the poverty items. We compared the scores between thetwo sets of responses. Results show overall acceptablelevel of reliability (over 0.7 for interclass correlation) forthe different poverty variables.

Quantitative variablesWe selected 17 indicators of deprivation reflectingaspects of well-being (table 1) identified by literaturereview and validated through FGDs with experts andPSMI/caregivers. Both groups identified and agreed ondeprivation cut-offs for each indicator through participa-tory deliberation.41 Some standard dimensions were notincluded due to lack of relevance in Delhi. For instance,few respondents lacked access to diet staples.1

We classified the selected indicators in three majordomains of deprivation: individual-level capabilities,household-level material well-being and individual-levelpsychosocial factors. The first domain was composed ofnine indicators. Access to secondary school was the indi-cator for education; dropping out before reaching sec-ondary school was the cut-off. Unemployment was amajor source of vulnerability; deprivation of work wasthe cut-off. Food security was measured by access tothree meals per day, and respondents eating less wereconsidered deprived. Following the Unicef definitions,improved indoor air quality using cooking gas, improveddrinking water by pipe into residence and improvedsanitation by private flush toilet defined absence ofdeprivation for indicators 6–8. Finally, individual incomeconstituted a monetary indicator.Material well-being of the household was composed of

two series of indicators. Three indicators outlined condi-tions of living: minimum space per person (deprivationthreshold of 40 square feet per person); home owner-ship (renting was the cut-off); and poor quality housingwas having either the flooring, walls or roof made ofKutcha (precarious or temporary) material. Materialwealth was defined by three complementary indicators:the household average per capita income (threshold atthe international poverty line of US$1.25 per day or 68Indian rupees);42 assets included typical goods owned bythe household;2 and monthly household expenditures.3

Finally, two psychosocial indicators were selected: phys-ical safety, measured through an indicator of perceptionof unsafe environment, and political participation in themunicipal elections.

1For vegan individuals, the diet staple included at least dal on a dailybasis; for non-vegan individuals, it included dairy products on a dailybasis. Meat for non-vegetarian individuals was not considered as a dietrequirement and therefore deprivation of meat is not an indicator ofpoor diet.2Assets include landline, mobile phones, wooden/steel sleeping cot,mattress, table, clock/watch, charpoy, refrigerator, radio/transistor,electric fan, television, bicycle, computer, moped/scooter/motorcycle,car.3Expenditures include food, health, school, transportation, savings andpersonal care products.

Trani J-F, et al. BMJ Open 2015;5:e006355. doi:10.1136/bmjopen-2014-006355 3

Open Access

Studies in India have shown that stigma resulting indiscriminatory practices is perceived to be high in thefamily and the community.43 44 As a result, we measuredexperienced discrimination as a dimension of stigmathrough self-evaluation of unfair treatment by the family.We asked all respondents whether they were excludedfrom family decision compared with other householdmembers of the same generation. Unfair treatmentwithin family is a feature of stigma in India.44 We testedthis through FGDs with PSMI of both genders. Wefound high association between SMI and exclusion fromregular family decisions, particularly for women.

Other dimensions of participation did not show anydiscriminatory process. Inclusion in community activitiesshowed similar 30% levels of participation betweenPSMI and controls. A possible explanation for participa-tion is that where symptoms of mental illness aremanaged by treatment family develop coping strategiesthrough symbolic social participation and selective dis-closure to avoid rejection, stigma and avoidance byothers associated with their relative’s condition.45–47

Finally, we enquired about participation in political activ-ities such as ‘gram sabhas’ or local associations. Wefound generalised low participation in political activities,

Table 1 Dimensions, indicators and cut-off of deprivation

Dimensions Indicators Questions Cut-off

Individual-level basic

capabilities

Health access Could you receive healthcare when sick? Deprived of healthcare

Education What is your level of education? Primary education

completed

Access to

employment

What is your usual primary activity? Not working

Food security How many meals are usually served in your

household in a day?

1 or 2 meals

Source of

drinking water

What is the primary source of drinking water? Pipe outside home/public

pump

tanker truck/cart with small

tank

water from a covered well

unprotected well

spring/river/dam/lake/pond/

stream

Indoor air quality What is the primary source of cooking fuel? Wood, coal/charcoal, dung,

kerosene, straw/shrubs/

grass/crop

Type of sanitation What type of toilet facilities do you use when at

home?

Open field, pit latrine

improved ventilated pit

public latrine

Type of lighting What is your primary source of lighting? Generator, kerosene lamp,

petromax, candle, none

Individual income What is your income? Less than $1.25 per day

Household-level

material well-being

Crowded space How many people live in the dwelling? Less than 50 square feet

per person

Housing

ownership

Does the family own the house? Do not own the house

Housing quality Are the material used for walls, floor and roof in

your house kutcha or pucca?

Any of walls, floor or roof is

kutcha

Assets ownership Do you possess any of the following? Mobile

phone, landline, wooden/steel sleeping cot,

mattress, table, clock/watch, charpoy, refrigerator,

radio/transistor, electric fan, television, bicycle,

computer, moped/scooter/motorcycle, car

Lowest two asset quintiles

Household per

capita income

What is the family income? Less than $1.25 per capita

per day

Household

expenditures

What is the household’s monthly expenditure? Less than $1.25 per capita

per day

Individual-level

psychosocial

dimensions

Physical safety How safe is the place where you live? Rather or very unsafe

Political

participation

Did you vote in the last municipal election? Did not vote

4 Trani J-F, et al. BMJ Open 2015;5:e006355. doi:10.1136/bmjopen-2014-006355

Open Access

which is a common feature in New Delhi and thereforenot a good indicator of experienced discrimination.

Statistical analysisOur primary aim was to explore the effect of mentalillness and stigma on poverty. We used an unmatchedMPI measure to identify differences in levels of povertybetween PSMI and controls.37 Dimensions were inde-pendently assessed and the method focused on dimen-sional shortfalls. This method allowed us to aggregatedimensions of multidimensional poverty measures andconsisted of two different forms of cut-offs: one for eachdimension and the other relating to cross-cutting dimen-sions. If an individual fell below the chosen cut-off on aparticular dimension, he/she was identified as deprived.The second poverty cut-off determined the number ofdimensions in which a person must be deprived to bedeemed multidimensionally poor.We first performed one-way analyses to assess differ-

ences in poverty levels and discrimination between PSMIand controls, by gender and caste adjusting for post hocpairwise comparisons using the Scheffe method. We alsocarried out correlation analysis to assess overlap ofdimensions of deprivation.We then calculated three indicators of multidimen-

sional poverty: (i) the headcount ratio (H), indicatinghow many people fall below each deprivation cut-off; (ii)the average poverty gap (A), denoting the averagenumber of deprivations each person experiences; and(iii) the adjusted headcount (M0), which is the head-count ratio (H) by the average poverty gap (A) and indi-cates the breadth of poverty. We established thecontribution of each dimension of poverty for cases andcontrols by dividing each of the two subgroups’ povertylevel by the overall poverty level, multiplied by the popu-lation portion of each subgroup.To assess potential bias in our estimates of MPI, we

carried out sensitivity analysis and compared three mea-sures of poverty with: (i) equal weight for every indicatorin each dimension, (ii) individual rankings of indicatorsdone by experts at Dr RML hospital during FGDs trans-formed into individual weights and then taking the

average of the individual weights,48 and (iii) groupranking based on the mean of individual rankings ofindicators during FGDs and taking the weight accordingto the group ranking.49 We found consistency acrossmeasures (data not shown).We finally calculated the crude and adjusted ORs with

associated 95% CIs using a logistic regression model toidentify association between stigma, SMI and multidimen-sional poverty. We used ‘no participation’ as the referencecategory. We defined a binary outcome for poverty (poor/non-poor) using the adjusted headcount ratio (M0) for acut-off k=6 corresponding to the highest gap betweenPSMI and controls. This cut-off corresponds to a preva-lence of poverty of 30.7% above the recent estimates of13.7% of urban Indians below the poverty line fixed at28.65 Indian rupees by the Indian Planning Commission,50

which has been criticised for being unrealistic. Thiscut-off is in line with World Bank recent estimate that33% of India’s population lives below the internationalpoverty line established at $1.25 per capita per day.51 Wecharacterised how SMI results in higher intensity ofmultidimensional poverty due to stigma. Aware thatstigma and discrimination may also affect women52 andmembers of lower castes,53 we adjusted the model forpotential confounders significantly associated withpoverty and family discrimination: caste (in case of dif-ference within the family), gender and age. We carriedout sensitivity analysis for different values of the cut-off kand found robustness in our model (data not shown).For all analyses, a p value of <0.05 was considered signifi-cant. Missing values were treated as being missing com-pletely at random. We used Stata (V.12.0) for databaseprocessing and all analysis.

RESULTSParticipantsWe interviewed 649 case patients and 647 controls. Ofthese, we excluded 110 (17%) cases and 151 (23%) con-trols, respectively, who did not complete the interview orfor whom the data were incomplete. The final analysisincluded 537 cases and 496 controls (figure 1). The

Figure 1 Flow chart depicting enrolment of patients with mental illness and controls without mental illness.

Trani J-F, et al. BMJ Open 2015;5:e006355. doi:10.1136/bmjopen-2014-006355 5

Open Access

Table 2 Characteristics of poverty and discrimination comparing patients and controls and by gender and caste

Dimension

PSMI

(n=647)

Control

(n=649) p Value

Male

PSMI

(n=411)

Male

controls

(n=408) p Value

Other

castes

PSMI

Other

castes

controls p Value

Female

PSMI

(n=238)

Female

controls

(n=238) p Value

ST/SC/

OBC

PSMI

ST/SC/

OBC

controls p Value

Health access 26 (4.0) 16 (2.9) 0.281 13 (3.2) 4 (1.0) 0.802 17 (4.8) 10 (2.5) 0.630 13 (5.5) 12 (5.0) 1.0 9 (3.3) 6 (2.5) 0.995

Education 155 (23.9) 129 (19.9) 0.086 70 (17.0) 52 (12.8) 0.511 61 (17.3) 59 (14.9) 0.879 85 (35.7) 77 (32.4) 0.843 82 (29.9) 65 (26.8) 0.850

Employment 396 (61.0) 252 (39.0) <0.0001 188 (45.7) 68 (16.7) <0.0001 222 (63.1) 151 (38.1) <0.0001 208 (87.4) 184 (77.3) <0.0001 164 (59.9) 96 (39.5) <0.0001

Food security 343 (52.9) 250 (38.6) 0.103 213 (51.8) 155 (38.0) 0.789 165 (46.9) 133 (33.6) 0.413 130 (54.6) 95 (39.9) 0.613 163 (59.5) 113 (46.5) 0.964

Source of water 122 (18.8) 118 (18.2) 0.724 86 (20.9) 74 (18.1) 0.732 62 (17.6) 61 (15.40) 0.881 36 (15.1) 44 (18.5) 0.837 55 (20.1) 56 (23.1) 0.893

Indoor air quality 48 (7.4) 38 (5.9) 0.271 35 (8.5) 24 (5.9) 0.515 17 (4.8) 13 (3.3) 0.861 13 (5.4) 14 (5.9) 0.998 27 (9.9) 24 (9.9) 1.0

Type of

sanitation

215 (33.1) 180 (27.8) 0.040 147 (35.8) 60 (25.2) 0.271 93 (26.4) 104 (26.3) 1.0 68 (28.6) 66.7 (29.4) 0.897 112 (40.9) 72 (29.6) 0.050

Type of lighting 7 (1.1) 10 (1.6) 0.458 4 (1.0) 8 (2.0) 0.674 0 (0) 4 (1.0) 0.675 3 (1.3) 2 (0.8) 0.984 6 (2.2) 6 (2.5) 0.994

Individual

income

369 (68.7) 238 (47.9) <0.0001 176 (53.3) 74 (24.3) <0.0001 199 (68.9) 138 (45.5) 0.932 193 (93.2) 164 (85.9) <0.0001 154 (68.1) 95 (52.8) 0.241

Crowded space 206 (31.7) 164 (25.4) 0.010 130 (32.0) 94 (23.3) 0.059 89 (25.3) 70 (17.7) 0.131 76 (32.3) 70 (29.7) 0.938 104 (38.0) 91 (37.5) 0.999

Housing

ownership

223 (41.5) 148 (29.8) <0.0001 160 (39.7) 119 (29.2) 0.028 152 (43.2) 75 (30.9) 0.002 99 (42.1) 78 (32.7) 0.264 99 (36.2) 119 (30.1) 0.667

Housing quality 39 (6.3) 13 (2.2) <0.0001 29 (7.1) 7 (1.67) 0.001 13 (3.7) 6 (1.5) 0.493 10 (4.2) 6 (2.5) 0.830 23 (8.4) 7 (2.9) 0.007

Assets

ownership

294 (45.3) 214 (33.1) <0.0001 201 (48.9) 125 (30.6) <0.0001 131 (37.2) 94 (23.7) 0.002 93 (39.1) 89 (37.4) 0.986 148 (54.0) 116 (47.7) 0.531

Household

income

287 (44.2) 239 (36.9) 0.002 176 (42.8) 142 (34.8) 0.082 132 (37.5) 116 (29.3) 0.096 111 (46.6) 97 (40.8) 0.553 141 (51.5) 119 (49.0) 0.907

Household

expenditures

373 (57.5) 393 (60.7) 0.978 238 (58.0) 239 (58.6) 0.799 180 (51.1) 209 (52.8) 0.947 135 (56.7) 154 (64.7) 0.571 178 (65.0) 180 (74.0) 0.4291

Physical safety 117 (18.0) 134 (20.7) 0.221 80 (19.6) 80 (19.6) 0.907 51 (14.5) 68 (17.2) 1.0 53 (22.3) 53 (22.3) 0.824 62 (22.6) 65 (26.8) 1.0

Political

participation

265 (40.8) 209 (32.3) 0.001 163 (39.7) 122 (29.9) 0.030 152 (43.2) 125 (31.6) 0.005 102 (42.9) 86 (36.1) 0.506 102 (37.2) 80 (32.9) 0.760

Discrimination in

family decisions

178 (27.4) 116 (17.9) <0.0001 71 (17.3) 12 (2.9) <0.0001 92 (26.1) 71 (17.9) 0.042 107 (45.0) 104 (43.7) 0.988 78 (28.5) 43 (17.7) 0.020

Missing values are missing completely at random and there was no significant statistical difference. Incidence of poverty expressed as a percentage is given in parentheses. All p values arecorrected for multiple comparisons using Scheffe method.OBC, other backward castes; PSMI, persons with severe mental illness; SC, scheduled castes; ST, scheduled tribes.

6TraniJ-F,etal.BM

JOpen

2015;5:e006355.doi:10.1136/bmjopen-2014-006355

OpenAccess

distribution between cases and controls was similar forgender (305 and 330 men, respectively, 61.5% in bothcases) and age (15–74 and 13–74 and median 35 and36, respectively).Table 2 reports the headcount ratios (H) or incidence

of deprivation in each dimension. There were statistic-ally significantly higher numbers of deprived PSMI thancontrols in nine dimensions. Differences were very highfor access to employment (28.1% difference), individualincome (20.7%) and relatively high for food security(15.1%) and house ownership (11.7%). In only onedimension—perception of physical safety—was there areverse non-significant difference as the number of con-trols was higher than the number of PSMI.Table 2 also show results by gender and caste.

Compared with male PSMI, the proportion of deprivedfemale PSMI was significantly higher (10 of 17 dimen-sions). Similarly, a higher number of PSMI (vs controls)from ‘scheduled castes’ (SC), ‘scheduled tribes’ (ST) or‘other backward castes’ (OBC) were poorer on 13 (vs 16dimensions) compared with PSMI (vs controls) fromunreserved castes.To investigate possible overlap of dimensions of

poverty, we calculated the estimates for the Spearman’srank correlation coefficients between each pair of dimen-sions of deprivation (table 3). We found no evidence ofstrong correlation between dimensions, illustratingabsence of association except for household income andexpenditures. We nevertheless kept both indicators to cal-culate MPI to account for information bias (particularlyrecall bias) often associated with measures of income inhousehold surveys.54 55 Significantly, this result demon-strates that a unique welfare indicator of poverty such asincome cannot represent all aspects of deprivation.

Multidimensional povertyResults in table 4 report the multidimensional headcountratio (H), the average deprivation shared across the poor(A) and the adjusted headcount ratio (M0) for all pos-sible cut-offs and for the two groups. Depending on thechosen cut-off, the proportion of PSMI and controls whowere multidimensionally poor varied greatly. For a cut-offof 1, 97.2% of PSMI and 91.7% of controls weredeprived: taking a union approach of deprivation in onedimension, this translates into quasi-universal poverty. Onaverage, PSMI were deprived on 5 dimensions andcontrols on 3.9. If multidimensional poverty requiresdeprivation in four, five or six dimensions simultaneously,the proportion of poor PSMI (compared with poor con-trols) becomes 68.5% (compared with 48.6%), 51.6%(35.9%) or 38.5% (22.2%). Conversely, if we adopt theintersection approach where being poor implies beingdeprived in all 17 dimensions, nobody in the sample ispoor and <1% of the sample is deprived in 13.The adjusted headcount ratio (M0) shows that PSMI

were worse off than controls for a cut-off (k) valuebetween 1 and 12 dimensions. This difference is signifi-cant (p<0.001) for (k)=1 to (k)=10 dimensions and

Table

3Spearm

an’s

correlationsbetweendim

ensions

Dim

ensions

Health

access

Education

Access

towork

Food

security

Sourceof

water

Indoorair

quality

Typeof

sanitation

Typeof

lighting

Individual

income

Crowded

space

Housing

ownership

Housing

quality

Assets

ownership

Household/

capital

income

Household

spending

Physical

safety

Political

participation

Healthaccess

1

Education

0.021

1

Accessto

work

0.1047*

0.1771*

1

Foodsecurity

0.0016

0.1309*

0.0878*

1

Sourceofwater

‐0.0277

0.1669*

0.0412

0.1263*

1

Indoorairquality

0.0341

0.1907*

0.0732*

0.1077*

0.1519*

1

Typeofsanitation

‐0.0103

0.1514*

0.0369

0.1045*

0.3026*

0.2440*

1

Typeoflighting

0.0193

0.0728*

0.0217

0.0642*

0.1079*

0.3018*

0.1550*

1

Individualincome

0.0801*

0.1865*

0.7373*

0.0788*

0.0534

0.0875*

0.0199

‐0.0134

1

Crowdedspace

‐0.0356

0.2471*

0.0521

0.1031*

0.1807*

0.1743*

0.2709*

0.0786*

0.0800*

1

Housingownership

0.0145

0.0138

0.029

0.0518

0.0553

‐0.0029

0.0207

0.0272

‐0.0123

0.1442*

1

Housingquality

0.0087

0.1739*

0.0764*

0.0558

0.2384*

0.2767*

0.3345*

0.0534

0.0824*

0.1969*

0.0182

1

Assets

ownership

0.0581

0.2727*

0.0751*

0.2544*

0.2364*

0.2820*

0.2330*

0.1634*

0.0797*

0.3079*

0.2926*

0.2753*

1

Household/capitaincome

0.0472

0.1949*

0.1623*

0.1513*

0.1989*

0.2070*

0.1597*

0.0805*

0.2066*

0.2712*

0.0443

0.1511*

0.2715*

1

Household

spending

0.0428

0.1667*

0.1062*

0.1483*

0.2377*

0.1568*

0.1409*

0.0760*

0.1381*

0.2792*

0.037

0.1533*

0.2331*

0.5360*

1

Physicalsafety

0.044

0.0406

0.0413

0.0596

0.1026*

0.0602

0.1223*

0.0609

0.0441

0.1723*

‐0.0252

0.0834*

0.0932*

0.1136*

0.1254*

1

Politicalparticipation

0.0188

‐0.0167

0.0386

0.0815*

0.1538*

0.031

0.1426*

0.0411

0.0125

0.1077*

0.2296*

0.0365

0.1617*

0.0714*

0.0735*

0.0493

1

*AllSpearm

ancorrelationcoefficients

significantatthe5%

levelorlower.

Trani J-F, et al. BMJ Open 2015;5:e006355. doi:10.1136/bmjopen-2014-006355 7

Open Access

highest (69% difference) for (k)=6. The average depriv-ation share (A) is higher among PSMI for a value of (k)between 1 and 5 and highest for (k)=1 (22% differ-ence). For a (k) between 6 and 14, the total number ofdeprivations faced by poor PSMI is slightly lower onaverage than for controls. Less than 30% of people werepoor in six dimensions or more, and the differencebetween PSMI and controls was the highest for a (k)value of 14 (7%).To further investigate the association between poverty

and mental illness, analysis was repeated for all possiblecut-offs and for gender and caste (table 4).Multidimensional poverty was significantly higher forfemale PSMI compared with female controls for any

threshold between one and seven dimensions (p<0.001)but also for male PSMI for any threshold between one andnine dimensions. On average, 62.8% of female PSMI weredeprived on five dimensions or more compared with35.9% of female controls, 44.5% of male PSMI and 25.6%of male controls. For female PSMI and controls—and malePSMI and controls, respectively—the difference is particu-larly pronounced and significant for highest cut-off values,and maximum for six and seven dimensions, respect-ively. The adjusted headcount ratio (M0) shows that SC/ST/OBC PSMI are worse off regardless of the value of(k) 1 through 10 than SC/ST/OBC controls and othercaste PSMI or controls. (M0) for SC/ST/OBC controlsis higher than for other caste PSMI for all (k) values.

Table 4 Multidimensional poverty measures for persons with severe mental illness (PSMI) and controls and by gender

and caste

Cut-off k

All PSMI Controls

T-value for M0‡

% difference

in M0*H† A M0 H A M0 H A M0

1 0.946 0.276 0.261 0.972 0.302 0.293 0.917 0.247 0.227 −6.574 29.3

2 0.849 0.301 0.256 0.901 0.321 0.289 0.792 0.277 0.219 −6.583 31.7

3 0.739 0.328 0.243 0.834 0.337 0.281 0.635 0.316 0.201 −7.051 39.9

4 0.590 0.367 0.216 0.685 0.372 0.255 0.486 0.359 0.175 −6.378 46.0

5 0.440 0.411 0.181 0.516 0.417 0.215 0.359 0.403 0.145 −5.210 48.5

6 0.307 0.462 0.142 0.385 0.458 0.177 0.222 0.471 0.104 −5.297 69.2

7 0.224 0.503 0.113 0.277 0.499 0.138 0.165 0.511 0.084 −4.062 64.0

8 0.144 0.553 0.080 0.175 0.550 0.096 0.111 0.559 0.062 −2.791 55.2

9 0.090 0.603 0.054 0.112 0.595 0.066 0.067 0.619 0.041 −2.334 61.6

10 0.055 0.650 0.036 0.069 0.636 0.044 0.040 0.676 0.027 −1.776 60.6

Cut-off k

PSMI Controls T-value

for M0

PSMI Controls T-value

for M0H M0 H M0 H M0 H M0

Female Male

1 0.990 0.327 0.917 0.227 −2.237 0.961 0.272 0.879 0.185 −6.7972 0.981 0.327 0.792 0.219 −2.322 0.852 0.265 0.702 0.175 −6.7173 0.942 0.322 0.635 0.201 −2.585 0.767 0.255 0.508 0.152 −7.1404 0.783 0.294 0.486 0.175 −2.157 0.624 0.230 0.364 0.127 −6.6525 0.628 0.257 0.359 0.145 −1.947 0.445 0.188 0.256 0.101 −5.3236 0.473 0.212 0.222 0.104 −2.191 0.330 0.154 0.148 0.069 −5.2637 0.343 0.166 0.165 0.084 −1.415 0.236 0.121 0.105 0.054 −4.3028 0.184 0.100 0.111 0.062 −0.396 0.170 0.094 0.079 0.043 −3.4389 0.116 0.068 0.067 0.041 −0.458 0.109 0.065 0.049 0.030 −2.75210 0.068 0.043 0.040 0.027 −0.157 0.070 0.044 0.030 0.019 −2.266

SC/ST/OBC Other castes

1 0.987 0.320 0.972 0.280 −2.437 0.958 0.264 0.884 0.194 −5.5322 0.942 0.317 0.900 0.276 −2.458 0.862 0.258 0.723 0.185 −5.5103 0.863 0.308 0.783 0.262 −2.496 0.799 0.251 0.545 0.164 −6.0974 0.748 0.288 0.628 0.235 −2.574 0.623 0.220 0.396 0.137 −5.2465 0.606 0.254 0.494 0.203 −2.262 0.426 0.174 0.274 0.109 −3.9276 0.460 0.211 0.306 0.148 −2.680 0.304 0.138 0.162 0.076 −3.8437 0.336 0.168 0.233 0.122 −1.917 0.215 0.106 0.125 0.063 −2.7888 0.217 0.118 0.161 0.092 −1.160 0.131 0.072 0.086 0.047 −1.8099 0.133 0.079 0.100 0.064 −0.757 0.090 0.053 0.050 0.030 −1.86410 0.075 0.048 0.061 0.043 −0.308 0.055 0.034 0.030 0.019 −1.459

Rows 11–17 are omitted as very few are deprived in >10 dimensions, no one is deprived in >15 dimensions.†H is the percentage of the population that is poor H=*(M0PSMI M0controls)/M0PSMI.‡ Adjusted Wald test for difference in adjusted headcount ratio between patients and controls. The average poverty gap (A) is not presentedfor gender and caste but can be easily calculated dividing the adjusted headcount (M0) by the headcount ratio (H).

8 Trani J-F, et al. BMJ Open 2015;5:e006355. doi:10.1136/bmjopen-2014-006355

Open Access

Table 5 presents the percentage contribution of eachdimension to (M0) for different (k). Deprivations inindividual income household expenditures and employ-ment were contributing each >10% to the overall (M0)for PSMI, whatever the value (k) between 1 and 8. Forcontrols, employment was a less salient contributorwhile the contribution from household income wasamong the highest.

Poverty and stigmaAssociation between multidimensional poverty andstigma was strong even when controlling for SMI, gender,caste and age (table 6; all p<0.0001). We included inter-action of stigma, SMI with caste and found that this termwas strongly and positively associated with a high level ofpoverty: the OR of being multidimensionally poor forPSMI from SC/ST/OBC compared with controls fromunreserved castes was 7.36 (95% CI 3.94 to 13.7).Similarly, we allowed for differential gender effects byincluding interaction of stigma and SMI with the genderof the respondent and found high effect on poverty:female PSMI were 9.61 (95% CI 5.58 to 16.5) more likelyto be poor compared with male controls.

DISCUSSIONOur findings establish that intensity of multidimensionalpoverty is higher for PSMI than the rest of the popula-tion. They also indicate that it is higher for women withSMI and for SC/ST/OBC with SMI. Deprivation ofemployment and income appear to be major contribut-ing factors to MPI. Lack of employment and incomeappears to aggravate mental illness. Finally, our findingssuggest that stigma linked to SMI, compounded withothers (particularly SC/ST/OBC and women), nega-tively impact poverty.The congruence of SMI and poverty, in a context of

high prejudice against mental illness, compromisesimprovement. Mental illness in India is linked to lack ofknowledge and pervasive negative assumptions, themost common being that PSMI are violent and unableto work.18 31 44 Not surprisingly, deprivation of employ-ment and income contributes highly to multidimen-sional poverty of PSMI compared with controls. Thisfinding ties in with a study on employment for Indianmen with schizophrenia, which found that employmentprovided not just an essential social role but was also acondition for rehabilitation, enhanced confidence andself-esteem.44

Although there is evidence of differences in mentalhealth outcomes between men and women, analyses ofgender disparities are lacking in literature on povertyand mental health in low-income countries.44 56 57 Inour sample, women with SMI were systematically moredeprived in higher numbers of dimensions. Similarly,SC/ST/OBC SMI–poverty associations were found to beconsistent across dimensions of poverty regardless of thethreshold for multidimensional poverty. These findings

Table

5Percentagecontributionofeachdim

ensionto

povertyforpersonswithsevere

mentalillness(PSMI)andcontrols

fork1–8

Cut-offk

Health

access

Levelof

education

Accessto

employment

Food

security

Sourceof

drinking

water

Indoorair

quality

Typeof

sanitation

Typeof

lighting

Individual

income

Crowded

space

Housing

ownership

Housing

quality

Assets

ownership

Household

income

Household

expenses

Physical

safety

Political

participation

1PSMI

0.86

4.63

11.62

10.87

3.74

1.31

4.37

0.22

13.79

4.56

8.33

1.27

4.82

8.86

12.63

3.59

4.52

Controls

0.78

5.33

7.74

10.15

4.86

1.57

3.50

0.47

12.45

4.71

7.74

0.58

3.92

9.62

16.42

5.75

4.39

2PSMI

0.87

4.70

11.79

10.58

3.75

1.33

4.32

0.23

13.91

4.62

8.04

1.29

4.89

8.95

12.62

3.60

4.51

Controls

0.76

5.41

8.00

9.68

5.03

1.62

3.62

0.49

12.43

4.86

7.19

0.59

4.05

9.95

16.43

5.57

4.32

3PSMI

0.86

4.79

11.77

10.44

3.86

1.36

4.29

0.23

13.64

4.72

8.07

1.33

4.99

9.00

12.51

3.55

4.60

Controls

0.77

5.61

8.15

9.33

5.31

1.77

3.54

0.47

12.16

5.14

6.85

0.65

4.43

10.45

15.58

5.55

4.25

4PSMI

0.95

4.94

11.05

10.49

3.91

1.46

4.47

0.26

12.77

4.99

7.78

1.46

5.42

9.42

12.55

3.57

4.51

Controls

0.68

5.77

7.95

8.83

5.57

1.90

4.01

0.54

11.14

5.57

6.66

0.75

4.82

10.80

14.95

5.50

4.55

5PSMI

0.87

5.25

10.30

10.24

4.33

1.63

4.59

0.31

11.67

5.30

7.54

1.68

6.17

9.73

12.18

3.72

4.49

Controls

0.74

6.39

7.79

8.36

5.90

2.05

4.26

0.66

10.49

6.15

6.48

0.90

5.33

10.82

13.77

5.41

4.51

6PSMI

0.99

5.46

9.86

9.99

4.65

1.86

5.09

0.25

11.10

5.58

6.95

2.05

6.45

9.74

11.85

3.66

4.47

Controls

0.80

7.05

7.27

7.50

6.59

2.73

4.66

0.91

9.32

7.05

6.59

1.25

6.36

9.89

12.05

5.57

4.43

7PSMI

1.11

5.62

9.65

9.57

4.91

2.14

5.22

0.32

10.76

5.70

6.41

2.37

7.04

9.57

11.16

3.88

4.59

Controls

0.42

7.44

7.02

7.16

6.60

3.37

5.06

1.12

9.13

7.72

6.18

1.40

6.74

9.55

11.38

5.62

4.07

8PSMI

0.91

5.23

8.65

9.22

5.46

2.62

5.80

0.34

9.90

6.37

7.05

2.73

7.96

8.76

10.35

3.98

4.66

Controls

0.38

7.07

6.12

6.88

6.88

4.21

5.93

1.34

8.22

7.65

6.12

1.91

7.65

9.56

10.52

5.54

4.02

Trani J-F, et al. BMJ Open 2015;5:e006355. doi:10.1136/bmjopen-2014-006355 9

Open Access

strongly suggest that stigma linked to various margina-lised groups have the power to accelerate and intensifyexclusion and related discrimination. For women, SMIcan negatively impact well-being in two ways. First, SMIlimits women from fulfilling family and social roles,leading to these women being considered a burden forthe family. This is true despite studies such as the Indianstudy of women with schizophrenia abandoned by theirhusbands who expressed the desire to work to supportthemselves.58 Second, traditional beliefs (punishmentfor previous lives, evil eye/curse), as well as negative layattitudes on causes and behaviours, lead to increaseddiscrimination of and sometimes violence against SMIs,particularly for women.59

Our study finds that SC/ST/OBC and poverty furthercompound SMI. Discrimination linked to caste in acces-sing education or employment has been a leitmotif inmodern India and only partially addressed through consti-tutional provisions and reservation policies. Caste discrim-ination still results in scant employment opportunities, lessaccess to secondary and higher education—key for sal-aried public and private jobs, perpetuating powerless-ness, traditional forms of dominance and oppression,inequalities, lower living standards among SC/ST/OBC as a entrenched social identity in India.60 61

This situation is even more catastrophic for PSMI fromSC/ST/OBC.It is clear that a ‘negative feedback loop’ exists:

stigma against SMI, particularly for SC/ST/OBC andwomen, is a strong predictor of persistent poverty.Moreover, stigma strongly bears on intensity of poverty.Stigma leads to difficulty for PSMI in finding andkeeping a job, and this also increases the perceivedburden of SMI by family members. In turn, this depriv-ation on various dimensions erodes self-esteem andbrings shame and acceptance of discriminatory atti-tudes.62 These compounding factors may result in aworsening of mental illness.Beyond the PSMI, stigma and discrimination have a

negative effect on family members and caregivers who oftenfeel ashamed, embarrassed or unable to cope with the

stigma.58 63–67 While there have been campaigns and pol-icies to address discrimination against SC/ST/OBC andwomen in India, no large-scale awareness campaign has everaddressed the prejudice and discrimination faced by PSMIs.This study has some limitations. First, a potential limita-

tion is that we measured experienced discrimination witha single-item question on exclusion from family decisionrather than a multiple-item scale. There was not a specificformalised psychometrically validated measure ofexperienced stigma available focusing on the scope andcontent of discrimination before the Discrimination andStigma Scale made available after our study was carriedout.10 Other factors may also explain exclusion fromfamily decisions, particularly symptomatic patients’ dis-ruptive behaviour. To account for this issue, we selecteda large sample of PSMI at Dr RML hospital representinga wide variety of severity of symptoms. Yet all outpatientswere successfully treated and mostly in follow-up, andtherefore not symptomatic at the time of the survey.Despite treatment, SMI in cases was significantly asso-ciated with our measure of stigma compared with con-trols, showing that ’pre-existing beliefs’ or stereotypeslinked to past experience with the mental illness werecritical to the activation of the discrimination processrather than the current mental health status of theperson.68 Second, it was not possible to establish the dir-ection of the association between poverty andSMI; poverty can be a cause as well as a consequence ofSMI. Third, SMI was diagnosed within a psychiatricdepartment of a free government hospital. Researchindicates the poorest members of society may still notaccess such services, even when free, possibly introdu-cing a selection bias in our sample.69 Additionally, PSMInot receiving medical treatment might be even moremarginalised, at greater risk of poverty than those receiv-ing healthcare. Thus the sampling bias might haveunderestimated association between SMI, stigma andpoverty. Finally, due to the large sample size we couldnot evaluate each control using detailed diagnostic psy-chiatric questionnaires but only screen them for majormental disorders.

Table 6 Logistic model for association between multidimensional poverty, stigma and severe mental illness (SMI)

Unadjusted model Adjusted model

OR 95% CI OR 95% CI

Family participation (no participation) 2.92 2.16 to 3.93 2.61 1.27 to 5.31

SMI (controls) 2.20 1.67 to 2.89 2.07 1.25 to 3.41

Women (men) 2.17 1.65 to 2.83 1.88 1.36 to 2.58

SC/ST/OBC (higher caste) 2.06 1.56 to 2.70 2.39 1.39 to 4.08

Age (in year) 0.99 0.97 to 0.99 0.98 0.96 to 0.99

Interaction terms

No participation×SMI (participation×controls) 6.38 3.49 to 11.6

No participation×SC/ST/OBC (participation×high caste) 4.86 2.19 to 10.7

No participation×women (participation×men) 4.63 2.60 to 8.21

No participation×women×SMI (participation×male×controls) 9.62 5.58 to 16.5

No participation×SC/ST/OBC×SMI (participation×high caste×controls) 7.36 3.94 to 13.7

SC scheduled castes; ST, scheduled tribes; OBC, other backward castes.

10 Trani J-F, et al. BMJ Open 2015;5:e006355. doi:10.1136/bmjopen-2014-006355

Open Access

CONCLUSIONOur study provides evidence that mental health profes-sionals must incorporate an understanding of multidi-mensional poverty stressors as well as address family andcommunity dynamics. Where resources are limited,medical professionals would benefit from working withpublic health and disability networks to weaken persist-ent stigma against SMI. Policies promoting employmentsupport for PSMI (notably through reservations or fairemployment policies, and access to credit) are criticallyimportant. The implications of our findings go beyondmedical and public health and link mental health tointernational development. Promoting employment andfighting social stigma for PSMI not only mitigates theimpact of illness for some but appears to be a centralconcern of global poverty.

Author affiliations1Brown School, Washington University in St. Louis, St. Louis, Missouri, USA2Program in Occupational Therapy, School of Medicine, WashingtonUniversity in St. Louis, St Louis, Missouri, USA3Department of Psychiatry & De-addiction Services, Resource Centre forTobacco Control, PGIMER- Dr. Ram Manohar Lohia Hospital, New Delhi, India4Nijmegen School of Management, Radboud University, Nijmegen,Netherlands5Leonard Cheshire Chair, Director, Leonard Cheshire Disability & InclusiveDevelopment Centre, Division of Epidemiology and Public Health UniversityCollege London, London, UK6UCL School of Life and Medical Sciences, University College London,London, UK

Contributors Study designed by JFT, SD, PB and SJ. Data collectionsupervised by SV, NM, SN and SD. Literature review by PB with JFT. Dataanalysis by JK and JFT. Data interpretation and writing by JFT and PB, andrevised by SD and NG. All authors contributed to the final manuscript. J-FT hadfull access to all data and takes final responsibility for publication submission.

Funding Funded by DFID through the Cross-Cutting Disability ResearchProgramme, Leonard Cheshire Disability and Inclusive Development Centre,University College London (GB-1-200474).

Competing interests None.

Ethics approval University College London Research Ethics Committee andthe Dr Ram Manohar Lohia Hospital Institutional Ethics Committee.

Provenance and peer review Not commissioned; externally peer reviewed.

Data sharing statement Extra data can be accessed via the Dryad digitalrepository at http://doi.org/10.5061/dryad.NNNNN with doi:10.5061/dryad.j167m.

Open Access This is an Open Access article distributed in accordance withthe Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license,which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, providedthe original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/

REFERENCES1. Lozano R, Naghavi M, Foreman K, et al. Global and regional

mortality from 235 causes of death for 20 age groups in 1990 and2010: a systematic analysis for the Global Burden of Disease Study2010. Lancet 2012;380:2095–128.

2. World Health Organization. Mental health and development: targetingpeople with mental health conditions as a vulnerable group. 2010.

3. Vos T, Flaxman AD, Naghavi M, et al. Years lived with disability(YLDs) for 1160 sequelae of 289 diseases and injuries 1990?2010:a systematic analysis for the Global Burden of Disease Study 2010.Lancet 2012;380:2163–96.

4. Whiteford HA, Degenhardt L, Rehm J, et al. Global burden of diseaseattributable to mental and substance use disorders: findings from theGlobal Burden of Disease Study 2010. Lancet 2013;382:1575–86.

5. Kleinman A, Hall-Clifford R. Stigma: a social, cultural and moralprocess. J Epidemiol Community Health 2009;63:418–19.

6. Boyce W, Raja S, Patranabish R, et al. Occupation, poverty andmental health improvement in Ghana. ALTER Eur J Disabil Res2009;3:233–44.

7. Elwan A. Poverty and disability: a survey of the literature.Washington, DC: Social Protection Advisory Service, 1999.

8. Groce N, Kett M, Lang R, et al. Disability and Poverty: the need for amore nuanced understanding of implications for development policyand practice. Third World Q 2011;32:1493–513.

9. Skeen S, Lund C, Kleintjes S, et al. Meeting the MillenniumDevelopment Goals in Sub-Saharan Africa: what about mentalhealth? Int Rev Psychiatry 2010;22:624–31.

10. Brohan E, Clement S, Rose D, et al. Development and psychometricevaluation of the Discrimination and Stigma Scale (DISC).Psychiatry Res 2013;208:33–40.

11. Draine J. Mental health, mental illnesses, poverty, justice, and socialjustice. Am J Psychiatr Rehabil 2013;16:87–90.

12. Fone D, Greene G, Farewell D, et al. Common mental disorders,neighbourhood income inequality and income deprivation:small-area multilevel analysis. Br J Psychiatry 2013;202:286–93.

13. Li N, Pang L, Du W, et al. Association between poverty andpsychiatric disability among Chinese population aged 15–64 years.Psychiatry Res 2012;200:917–20.

14. Chatterjee P. Economic crisis highlights mental health issues inIndia. Lancet 2009;373:1160–1.

15. Kuruvilla A, Jacob KS. Poverty, social stress & mental health. IndianJ Med Res 2007;126:273–8.

16. Das J, Do QT, Friedman J, et al. Mental health and poverty indeveloping countries: Revisiting the relationship. Soc Sci Med2007;65:467–80.

17. de Menil V, Osei A, Douptcheva N, et al. Symptoms of commonmental disorders and their correlates among women in Accra,Ghana: a population-based survey. Ghana Med J 2012;46:95–103.

18. Link BG, Phelan JC. Conceptualizing stigma. Annu Rev Sociol2001;27:363–85.

19. Thornicroft G, Rose D, Kassam A, et al. Stigma: ignorance,prejudice or discrimination? Br J Psychiatry 2007;190:192–3.

20. Ilic M, Reinecke J, Bohner G, et al. Belittled, avoided, ignored,denied: assessing forms and consequences of stigma experiencesof people with mental illness. Basic Appl Soc Psychol2013;35:31–40.

21. Wing JK, Agrawal N. Concepts and classification of schizophrenia.In: Hirsch SR, Weinberger DR. eds. Schizophrenia. Oxford:Wiley-Blackwell, 2003:3–15.

22. Corrigan PW, Markowitz FE, Watson AC. Structural levels ofmental illness stigma and discrimination. Schizophr Bull2004;30:481–91.

23. Link BG, Struening EL, Neese-Todd S, et al. The consequences ofstigma for the self-esteem of people with mental illnesses. PsychiatrServ 2001;52:1621–6.

24. Watson AC, Corrigan P, Larson JE, et al. Self-stigma in people withmental illness. Schizophr Bull 2007;33:1312–18.

25. Rüsch N, Angermeyer MC, Corrigan PW. Mental illness stigma:concepts, consequences, and initiatives to reduce stigma.Eur Psychiatry 2005;20:529–39.

26. Wahl OF. Stigma as a barrier to recovery from mental illness.Trends Cogn Sci 2012;16:9–10.

27. Link BG, Phelan JC. Stigma and its public health implications.Lancet 2006;367:528–9.

28. Link BG, Yang LH, Phelan JC, et al. Measuring mental illnessstigma. Schizophr Bull 2004;30:511–41.

29. Hatzenbuehler ML, Phelan JC, Link BG. Stigma as a fundamentalcause of population health inequalities. Am J Public Health2013;103:813–21.

30. Crabb J, Stewart RC, Kokota D, et al. Attitudes towards mentalillness in Malawi: a cross-sectional survey. BMC Public Health2012;12:541.

31. Gureje O, Lasebikan VO, Ephraim-Oluwanuga O, et al. Communitystudy of knowledge of and attitude to mental illness in Nigeria. Br JPsychiatry 2005;186:436–41.

32. Jadhav S, Littlewood R, Ryder AG, et al. Stigmatization of severemental illness in India: against the simple industrializationhypothesis. Indian J Psychiatry 2007;49:189–94.

33. Ngui E, Khasakhala A, Ndetei D, et al. Mental disorders, healthinequalities and ethics: a global perspective. Int Rev Psychiatry2010;22:235–44.

Trani J-F, et al. BMJ Open 2015;5:e006355. doi:10.1136/bmjopen-2014-006355 11

Open Access

34. Ssebunnya J, Kigozi F, Lund C, et al. Stakeholder perceptions ofmental health stigma and poverty in Uganda. BMC Int Health HumRights 2009;9:5.

35. Phelan JC. Genetic bases of mental illness-A cure for stigma?Trends Neurosci 2002;25:430–1.

36. Khandelwal SK, Pattanayak RD. Fight against Stigma. In: ChavanBS, Gupta N, Arun P, Sidana AK, Jadhav S. eds. Comprehensivetextbook on community psychiatry in India. Delhi: Jaypee Brothers,2012:334–44.

37. Alkire S, Foster J. Counting and multidimensional povertymeasurement. J Public Econ 2011;95:476.

38. World Health Organization. ICD-10 classification of mental andbehavioral disorders: clinical descriptions and diagnostic guidelines.Geneva: World Health Organisation, 1992.

39. Dupont WD, Plummer WD Jr. Power and sample size calculations.A review and computer program. Controlled Clin Trials1990;11:116–28.

40. DeVellis RF. Scale development: theory and applications. Vol 26.Thousand Oaks: Sage, 2012.

41. Sen AK. Capabilities, lists, and public reason: continuing theconversation. Feminist Econ 2004;10:77–80.

42. Ravallion M, Chen S, Sangraula P. Dollar a day revisited.World Bank Econ Rev 2009;23:163–84.

43. Shrivastava A, Johnston M, Thakar M, et al. Origin and impact ofstigma and discrimination in schizophrenia—patients’ perception:Mumbai study. Stigma Res Action 2011;1:67–72.

44. Loganathan S, Murthy RS. Living with schizophrenia in India: genderperspectives. Transcult Psychiatry 2011;48:569–84.

45. Corrigan PW, Miller FE. Shame, blame, and contamination: a reviewof the impact of mental illness stigma on family members. J MentHealth 2004;13:537–48.

46. Karnieli-Miller O, Perlick DA, Nelson A, et al. Family members’ ofpersons living with a serious mental illness: experiences and effortsto cope with stigma. J Ment Health 2013;22:254–62.

47. Larson JE, Corrigan P. The stigma of families with mental illness.Acad Psychiatry 2008;32:87–91.

48. Bruckner T, Catalano RA. Economic antecedents of sudden infantdeath syndrome. Ann Epidemiol 2006;16:415–22.

49. Eskenazi B, Marks AR, Catalano R, et al. Low birthweight inNew York city and upstate New York following the events ofSeptember 11th. Hum Reprod 2007;22:3013–20.

50. Planning Commission. Poverty estimates for social groups: 2004–05and 2011–12. New Delhi, 2013.

51. Olinto P, Beegle K, Sobrado C, et al. The State of the Poor: whereare the poor, where is extreme poverty harder to end, and what isthe current profile of the world’s poor? Econ Premise 2013;125:1–8.

52. Kohler Riessman C. Stigma and everyday resistance practices:childless women in South India. Gender Soc 2000;14:111–35.

53. Jaspal R. Caste, social stigma and identity processes. Psychol DevSoc 2011;23:27–62.

54. Biemer PP. ed. Measurement errors in surveys. New York: JohnWiley and Sons, 1991.

55. Biemer PP, Lyberg LE. Introduction to survey quality. Hoboken: JohnWiley & Sons, 2003.

56. Das J, Das RK, Das V. The mental health gender-gap in urbanIndia: patterns and narratives. Soc Sci Med 2012;75:1660–72.

57. Das J, Quy-Toan Do Q, Friedman J, et al. Mental health and povertyin developing countries: revisiting the relationship. Soc Sci Med2007;65:467–80.

58. Thara R, Kamath S, Kumar S. Women with schizophrenia andbroken marriages—Doubly disadvantaged? Part I: Patientperspective. Int J Soc Psychiatry 2003;49:225–32.

59. Human Rights Watch. “Treated Worse than Animals”. Abusesagainst women and girls with psychosocial or intellectual disabilitiesin institutions in India. Human Rights Watch, 2014.

60. Jeffrey C, Jeffery P, Jeffery R. Reproducing difference? Schooling,jobs, and empowerment in Uttar Pradesh, India. World Dev2005;33:2085–101.

61. Kijima Y. Caste and tribe inequality: evidence from India,1983–1999. Econ Dev Cult Change 2006;54:369–404.

62. Corrigan PW, Larson JE, Rüsch N. Self-stigma and the “why try”effect: Impact on life goals and evidence-based practices. WorldPsychiatry 2009;8:75–81.

63. Thara R, Kamath S, Kumar S. Women with schizophrenia andbroken marriages—Doubly disadvantaged? Part II: Familyperspective. Int J Soc Psychiatry 2003;49:233–40.

64. Brady N, McCain GC. Living with schizophrenia: a familyperspective. Online J Issues Nurs 2005;10:7.

65. Lee S, Lee MTY, Chiu MYL, et al. Experience of social stigma bypeople with schizophrenia in Hong Kong. Br J Psychiatry2005;186:153–7.

66. Yang LH, Purdie-Vaughns V, Kotabe H, et al. Culture, threat, andmental illness stigma: identifying culture-specific threat amongChinese-American groups. Soc Sci Med 2013;88:56–67.

67. Yang LH, Pearson VJ. Understanding families in their own context:schizophrenia and structural family therapy in Beijing. J Fam Ther2002;24:233–57.

68. Link BG, Cullen F, Frank J, et al. The social rejection of formermental patients: understanding why labels matter. AJS1987;92:1461–500.

69. Murali V, Oyebode F. Poverty, social inequality and mental health.Adv Psychiatr Treat 2004;10:216–24.

12 Trani J-F, et al. BMJ Open 2015;5:e006355. doi:10.1136/bmjopen-2014-006355

Open Access